The UAE-based Interactive Robots and Media Lab (IRML) engaged in the creation of the world’s first Arabic speaking and comprehending android by the name of Ibn Sina. Its founder and director Nikolaos Mavridis, PhD, Massachusetts Institute of Technology (MIT) and one of the leading experts in the field of Artificial Intelligence (AI) was interviewed by researchers from the Access to Knowledge for Development Center (A2K4D) of the American University in Cairo (AUC) School of Business, focusing on three key research areas: data, AI and development. In his interview with A2K4D, Mavridis gives an overview on how with the right combination of knowledge, empirical data, human resources and machines, along with the right ethical principles, and appropriate institutions, AI can create a better world. Responses have been edited for clarity and brevity.
“..the future is humans AND machines; not humans OR machines,” states Nikolaos Mavridis, PhD, MIT; one of the creators behind the first Arabic speaking robot.
Tell us how AI is currently being used in the UAE and MENA.
Today, a big part of the everyday narrative in MENA has to do with AI, robotics and drones. Many related activities happen regularly: competitions and government-funded initiatives with prizes, conferences and summer schools. For example, the “Drones for Good” and “Robotics and AI for good” competitions that the government of Dubai had organized, and in which global companies and institutions took part with the aim of promoting socially beneficial uses of these technologies.
Do you think MENA countries are producers, or adaptors of AI?
At this moment, we are primarily consumers and adopters of AI, and have just started to be producers, yet at a small scale. We are, however, increasingly large producers of data, but not of codes and algorithms. Innovation does exist in applying such technologies in MENA; but not yet in creating them, and this presents an opportunity for the region to reduce dependence on external entities and enhance capacities of a pool of trusted experts who can deal with these technologies.
However, an important question persists: how can MENA countries also become producers, and not only consumers, of such advanced technologies? One of the things that is missing is trust for local innovation. Many businessmen and investors do not believe that anything worthwhile could ever be produced inside MENA; and they just find the “safe solutions” that have first happened in the west. Think, for example, of two of the big success stories of companies from the region: Souq.com and Careem. Was the technology new and innovative? Not really! They were essentially Amazon and Uber. Such examples illustrate that, although they make business sense, still not much innovation is produced in our region. And things should change. And the right kind of role models, of companies really producing innovative solutions that have started in the MENA region, could play a very important role in that respect.
What are some examples of AI production and some challenges around it?
Traditionally, there have been application of “wider-sense” AI in a few fields; for example, in airports and airline scheduling and dynamic rescheduling systems. Another more recent application is credit risk assessment systems for banks that need to take an informed decision before administering a loan. Quite importantly, computer-based surveillance systems are becoming increasingly important; and in countries such as China, where they are widely deployed, they provide security as well as optimizations of many organizational and behavioral aspects.
And of course, for the case of robotics, applications in manufacturing have existed for decades; but now the so-called co-robots are starting to work alongside humans in the production lines. Also, notably, many different application areas are starting to become populated by robots: medical robots, military robots, educational robots, search-and-rescue robots; as well as many other kinds of service robots. Classic examples that have attracted the media in the Gulf include the “policeman” robots that were shown to be wandering around Dubai Mall; and of course, the announcements about the first experimental flights of Drone Taxis in the Emirates.
In relation to this, the theory proposed by Daron Acemoglu from the Center for the Digital Economy at MIT regarding technological unemployment, states that historically the entry of robots in manufacturing would result to roughly 5 human jobs being lost for every new robot entering a plant, and that -even worse- the salaries of the remaining humans would also decrease. But, as Daron proposed, if production lines are re-designed so that instead of robots replacing humans, the demands of each production cell are increased, this creates a need for humans to be working alongside robots. If humans carry out the right complementary tasks in which they are better suited than robots, then the prediction is that not only jobs will not be lost, but more human jobs will be created; and interestingly enough, we might even experience an increase in human salaries. This, though, still remains a theory, and needs to be tested and proved in practice. One should remember: the future is humans AND machines; not humans OR machines.
And then, there is the issue of “biases” in AI. A classic example is in supervised machine learning where a system is trained based on a so-called “training set”. But such training sets, are often created by humans or minded through humanly created data: for example, opinions posted on the internet. Such humanly created data though often come with biases. One example is gender biases that can be transferred to machine-learning-driven Human Resources (HR) decision-support systems used for deciding which candidate to hire for a position. In that sense, the undesirable biases of the humanly created training set that was used to train the AI, are carried over (and in some cases even magnified!) to the AI system.
Do data limitations influence the current state of AI in MENA?
Data is a prerequisite for data-driven models that comprise an important part of AI techniques. The magnitude of data required for the functioning of AI-based systems depends on the case. There are cases where a small and easily obtainable amount of data can suffice for a near-optimal performance of the system being built; and in this case further data offers no real advantage.
Certainly though, specific types of data, are becoming increasingly important for the fine-tuned operation of many of our modern-day systems that support our activities, our businesses, our cities and nations.
Are you concerned with data ethics and privacy?
Generally speaking, people are usually highly unaware, and also very busy to understand the value of the information that they are handing out for free, and what can be done with this information. Therefore, not only appropriate awareness should increase, but also the tools that will enable the average citizen to monitor what information he or she is giving out and how it could be used, to select what to give to whom, and to also be able to ask for compensation (in various forms; services, monetary, etc.), in return to what they are giving out. The monetization of people’s data, and the ethical and unethical sides of it, is an important conversation.
And those that are more involved with the design, creation, and application of AI and the other novel technologies of the fourth industrial revolution, should be even more concerned with data ethics, as compared to the average citizen. That entails being aware of the fundamental issues and their potential implications, as well as of the methods available for resolving problems, current legislation, and the institutions and organizations that are related to technology ethics and to data ethics in particular. Furthermore, with rapid advances in the field emerges the need for frequent updates on the relevant knowledge regarding ethics and privacy.
What would be an enabling environment for AI development?
Legislation – and rather applicable legislation- is very important. Without it, certain technologies cannot be properly deployed, such as drones for example. Another important requirement is the readiness of public opinion; the ecosystem of opinions and attitudes of different stakeholder groups that are involved in the application of a specific technology. For example, for the case of robots to help the elderly, the opinions and attitudes of the elderly themselves, and all the directly and indirectly relevant stakeholder groups should shape the solution.
Human capital is another important area where appropriate interdisciplinarity is also needed: it is not enough to have technologists only. We also need people who can predict, measure, and regulate the psychological and socio-economic effects of new technologies, and you need appropriate decision makers at the governmental level, often aided by teams of complementary experts. Furthermore, you need the appropriate mass media entities, that can introduce relevant ideas and role models – going beyond the globally visible models such as Bill Gates, Elon Musk, and [Mark] Zuckerberg. You also need the right kind of public intellectuals, that can provide credible narratives to citizens.
What is the role of government in supporting AI for development in the MENA region?
The role of government can be quite crucial; although sometimes it is not adequate on its own, without the support from external entities, especially in the case of poorer countries. International NGOs and non-profits, humanitarian and even foreign government-connected aid organizations, can also play a sometimes vital role towards supporting AI for inclusion and development. Also, appropriate coordination with private entities, that can direct monetary or in-kind resources, as a form of Corporate Social Responsibility, can also be crucial in some cases.
This article is based on an interview conducted by A2K4D researchers Farah Ghazal and Hana Shaltout for upcoming A2K4D research on the state of AI in the MENA. This article and research are based on a forthcoming chapter by A2K4D director, Professor Nagla Rizk. [Rizk, N. (forthcoming May 2020). “Artificial Intelligence and Inequality in the Middle East: The Political Economy of Inclusion” In Oxford Handbook of Ethics of Artificial Intelligence”, M Dubber, F Pasquale and Sunit Das, eds. Oxford: Oxford University Press]
